Maglev and abort_transaction-ing on every request

The code:

Running vanilla Rack::Lobster:

Requests per second: 425.62 [#/sec] (mean) Time per request: 2.350 [ms] (mean) Time per request: 2.350 [ms] (mean, across all concurrent requests)

Running Rack::Lobster and Maglev#abort_transaction-ing with each request:

Requests per second: 404.73 [#/sec] (mean) Time per request: 2.471 [ms] (mean) Time per request: 2.471 [ms] (mean, across all concurrent requests)

These are the best performances from several runs. Difference is pretty much negligible. Certainly varies enough to make the argument that the simplicity makes it worth it.